Optical information processing entails the ability to perform analog operations such as correlation, convolution, and differentiation with optical beams. These operations provide results that are both positive and negative in value. Since the intensity of an optical beam is always positive, positive and negative values may be represented as the difference in power between two optical beams. As a result, a signal initially in unipolar form (i.e., a signal represented by the always positive-valued intensity of a single optical beam) must be converted to a signal in bipolar form (i.e., a signal that has both positive and negative values). After processing the optical signal in bipolar form, it is often desirable to convert the result back into unipolar form, particularly when the result is intended to form an image. An output that forms an image is desirable because often it may be readily understood from a visual inspection.
An apparatus for converting an optical bipolar signal to a unipolar signal is described by Miller in IEEE J. Quantum Electron., Vol. 29, Number 2, February 1993, pages 678-698, specifically on page 681 and in FIG. 3 of that article. One limitation of this apparatus is that an optical bias beam must be added to one of the rails of the bipolar input signal. Accordingly, this known apparatus requires a relatively complex optical system.